autom8ic.
01 Case study Lead research

A research team,
not a research bot.

Most "AI lead research" tools are one language model holding too many jobs at once. This is four specialist agents collaborating on a single brief — and you can read what they produced right below.

Production · live 4-agent architecture ~£0.04 per lead 60s end-to-end
02 · THE BRIEF, UNEDITED

Generated in 58 seconds from one input.

Single input: https://posthog.com. No edits, no human in the loop. Read it the way a sales lead would.

Output · unedited
Lead Brief · PostHog · 12 May 2026

#Company Overview

PostHog provides an all-in-one product development platform called "Product OS" that combines product analytics, session replay, feature flags, a data warehouse, and AI-powered automation tools. Their target customers are product engineers and development teams at tech companies who need to understand user behaviour and iterate quickly. The company uses usage-based pricing with generous free tiers and positions itself with an irreverent, anti-corporate brand tone that emphasises engineering credibility and transparent operations.

#What's Happening Right Now

PostHog raised a $75 million Series E at a $1.4 billion valuation on September 29, 2025, led by Peak XV Partners. The company launched PostHog Code in beta approximately seven days ago, a desktop application that runs coding agents on top of product data. On September 10, 2025, PostHog redesigned its website to function like an operating system, though it discontinued its AI product assistant feature on November 19, 2025.

#Customer Base & Social Proof

PostHog maintains a customers page and case study documentation, but specific named customers and detailed testimonials are not publicly evident from search results. The available information points to customer infrastructure and case study frameworks rather than displaying concrete client examples.

#Outreach Angle

Given PostHog's recent $75 million Series E and the beta launch of PostHog Code, the company is clearly in aggressive growth mode with focus on AI-driven product development tools. Outreach should acknowledge their engineering-first culture and anti-corporate positioning — avoid generic SaaS pitch language and instead lead with technical specifics about how your solution integrates with their Product OS stack. Reference their transparent approach to operations when proposing partnerships, as this aligns with their stated brand values.

03 · HOW IT WORKS

Four agents. Each with one job.

Instead of one Claude call doing everything, three specialists run in parallel — each reading one source — and a fourth synthesises their findings into the final brief. Bounded scope per agent means no hallucinated cross-references, no infinite loops, and predictable per-lead cost.

  [ Form submission ]
         
         ├──→ Fetch Homepage ──→ Clean HTML ──┐
         ├──→ Search News ─────────────────────┤
         └──→ Search Customers ────────────────┤
                                                  
                              [ Specialist 1 ]  [ Specialist 2 ]  [ Specialist 3 ]
                                 Homepage          News            Customers
                                                  
                                          [ Synthesizer ]
                                                  
                                            Final brief
01
Bounded scope
Each specialist reads one source and answers four questions. No mixing, no loops, no hallucinated cross-references.
02
60× cheaper input
Custom HTML preprocessor strips scripts and tags before any AI sees the page. Typical homepage: 257k chars → 4k chars.
03
Built for production
Self-hosted. Error workflow attached. Retry layer on every external call. Runs without you watching.
04 · UNDER THE HOOD

The stack, plainly stated.

Orchestration
n8n, self-hosted on Hetzner
Language model
Claude Sonnet 4.5 via Anthropic API
Web search
Serper (Google search API)
Infrastructure
Docker, Caddy, HTTPS via Let's Encrypt
Error handling
Global error workflow + retry layer
Your data
Stays on your infrastructure. Always.
05 · GET IN TOUCH

Want one of these
built for your team?

I build production-grade AI automation systems for teams that want their workflows to actually run — not just demo well. Drop me a line.

hello@autom8ic.com →